Function equal_width_binning

src/memory/processing/hvg/mod.rs:219–273

Create equal-width bins for gene expression levels. Divides the expression range into equal-width bins for the Seurat method. This allows modeling of mean-variance relationship by expression level, as highly expressed genes naturally have different variance characteristics. ## Parameters `log_means` - Log-transformed mean expression values `n_bins` - Number of bins to create ## Returns Bin assi

(log_means: &[f64], n_bins: usize)

/// 3. Assign each gene to appropriate bin
/// 4. Handle edge cases (NaN values, exact maximum)
fn equal_width_binning(log_means: &[f64], n_bins: usize) -> anyhow::Result<(Vec<usize>, Vec<f64>)> {
    // Find min and max (excluding NaN)
    let mut valid_means: Vec<f64> = log_means
        .iter()
        .filter(|x| x.is_finite())
        .copied()
        .collect();

    if valid_means.is_empty() {
        return Err(anyhow::anyhow!("No valid mean values found"));
    }

    valid_means.sort_by(|a, b| a.partial_cmp(b).unwrap());
    let min_mean = valid_means[0];
    let max_mean = valid_means[valid_means.len() - 1];

    // Create equal-width bins
    let bin_width = (max_mean - min_mean) / n_bins as f64;
    let mut bin_edges = vec![0.0; n_bins + 1];

    for (i, edge) in bin_edges.iter_mut().enumerate().take(n_bins + 1) {
        *edge = min_mean + (i as f64) * bin_width;
    }

    // Make sure the last edge includes the maximum value
    bin_edges[n_bins] = max_mean + 1e-10;

    // Assign each gene to a bin
    let mut bin_indices = vec![0; log_means.len()];

    for (i, &mean) in log_means.iter().enumerate() {
        if !mean.is_finite() {
            bin_indices[i] = 0; // Assign NaN/inf to first bin
            continue;
        }

        // Find which bin this value belongs to
        let mut bin_idx = 0;
        for j in 0..n_bins {
            if mean >= bin_edges[j] && mean < bin_edges[j + 1] {
                bin_idx = j;
                break;
            }
        }

        // Handle edge case where mean == max_mean
        if mean == max_mean {
            bin_idx = n_bins - 1;
        }

        bin_indices[i] = bin_idx;
    }

    Ok((bin_indices, bin_edges))
}

/// Calculate mean and standard deviation of dispersions within each expression bin.
///